Murdoch University Research Repository

Welcome to the Murdoch University Research Repository

The Murdoch University Research Repository is an open access digital collection of research
created by Murdoch University staff, researchers and postgraduate students.

Learn more

Real-time prediction model for indoor temperature in a commercial building

Afroz, Z., Urmee, T., Shafiullah, GM.ORCID: 0000-0002-2211-184X and Higgins, G. (2018) Real-time prediction model for indoor temperature in a commercial building. Applied Energy, 231 . pp. 29-53.

Link to Published Version:
*Subscription may be required


Indoor environmental parameters have significant influence on commercial building energy consumption and indoor thermal comfort. Prediction of these parameters, especially that of indoor air temperature, along with continuous monitoring and control of real world parameters can aid in the management of energy consumption and thermal comfort levels in existing buildings. An accurate indoor temperature prediction model assists in achieving an effective energy management strategy such as resetting air temperature set-points in commercial buildings. This study examines the real indoor environmental data for multiple adjacent zones in a commercial building in the context of thermal comfort and identifies the possibility of resetting air temperature set-point without compromising the occupant comfort level. Also, the value of predicting the indoor temperature accurately in such a building is established through this case study. This study presents a nonlinear autoregressive network with exogenous inputs-based system identification method to predict indoor temperature. During model development efforts have been paid to optimize the performance of the model in terms of complexity, prediction results and ease of application to a real system. The performance of single-zone and multi-zone prediction models is evaluated using different combinations and sizes of training data-sets. This study confirms that evaluating the performance of the model in the context of major contributing aspects such as optimal input parameters and network size, optimum size of training data, etc. offers optimized model performance. Thus, when the developed model is used for long-term prediction, it provides better prediction performance for an extended time span compared to existing studies. Therefore, it is anticipated that implementation of this long-term prediction model will offer greater energy savings and improved indoor environmental conditions through optimizing the set-point temperatures.

Item Type: Journal Article
Murdoch Affiliation(s): School of Engineering and Information Technology
Publisher: Elsevier B.V.
Copyright: © 2018 Elsevier Ltd.
Item Control Page Item Control Page